• Canoe 2011 Decentralization

Decentralizing online social networks - just an academic exercise?

Participants:

• Kaiwen
• Navneet
• Chen Chen
• Vinay
• Zhang

Guiding Questions

• What aspects of social networks lend themselves to decentralization?
• Why would this make sense and what are the challenges to overcome?

Summary of breakout session discussion

Motivation for decentralization:

• Decentralization to eliminate central authority
  • Privacy
  • Control over data
  • Democratic model
• Decentralized social networks can be immortal and immune to restrictions
  • Facebook, Twitter ban in China and Egypt banning social networks during crisis

  • Service providers can die out and leave users with no service and data even worse misuse of data, eg: MySpace death

• Technical
  • Scalability
  • Load balancing?

Existing Decentralized OSNs

• Diaspora
  • Diaspora works by letting users set up their own server (or "pod") to host content; pods can then interact to share status updates, photographs, and other social data with encryption

• PeerSon

  • P2P effort to decentralize OSN

Challenges and Solutions for decentralization

• Decentralization of credentials

  • Where to store user credentials? (open problem)

  • granularity of credentials
    • per user per item
    • Certificate per session, issued by user or a trusted friend on user's behalf

    • Can we learn something from BitCoin? The way they prove the ownership of coins (Public key cryptography)

• Search and Recommendations
  • Search profiles/data
  • Link predictions / friend recommendations
• Group Communication / notification system
  • Pub/Sub system? But no persistence and availability

  • How to implement a decentralized Thialfi like system? (open problem)

• Personal/Private communication
• Integration with/migration from “traditional” Social Networks
• Scalability
  • How to scale to millions of users in a decentralized OSN?
• Decentralized storage
  • Decentralized partitioning

    • How to do efficient graph partitioning like SPAR in decentralized way? (open problem)

  • Data is already partitioned
    • assumes each user makes independent decision for storage
  • Most systems assume storage in DHTs
    • Are DHTs ideal for storage of such large amount of data?
  • Can we make use of social relations to improve DHTs?
  • Should the storage be independent of social relations?
    • Friends have natural incentive of hosting friends data and provide their bandwidth for hosting data
    • Should we use both techniques to achieve durability and availability of stored data?

Business Models

• Traditional business model
  • 'free' and reliable service at the expense of privacy
• Are users willing to live with lower quality of service in order to preserve their privacy?
  • Or to pay the required price to preserve quality of service?
• Can a decentralized advertisement platform generate revenue for users such that this revenue can pay for a more reliable decentralized social network infrastructure?